Can you explain to me what this ticket is about:
http://trac.dcache.org/trac.cgi/ticket/118
I think its a mislabeling.
The startup scripts make no such assumption. People have been able to
deploy PNFS on a different host for years and do so. I guess most Tier 1
sites do so.
I agree.
I suspect you are confused by the variable called ADMIN_NODE in
node_config, which points to the PNFS server, but that has always been
the case (always = at least the last one and a half years). The variable
is not used for anything else but to identify the PNFS server. Owen and
I already talked about this, but Tigran/Patrick can you please confirm
that this ticket should be closed - I do not see a bug; only a badly
named variable.
Cheers,
/gerd
I imagine that my fellow developers at desy where in to much of a hurry
to look into it, its not unusual for bugs to be mislabeled.
05/06/08 06:57:01 changed by patrick
* milestone changed from 1.8.0-15 to 1.8.0-15p3.
No this is unfortunately untrue, But makes little difference. Since it
was never the nature of the bug it was only a misunderstanding.
I 100% agree with you Gerd. This is a badly labeled variable that leads
to great confusion.
It will be out for the next major release, (relabeled as
"NAMESPACE_NODE" ) but I need to make it fully tested and 100% solid
before I change anything here, particularly when deployment bugs are
always of equal significance to data loss bugs. I have made all the
code effected work with either "NAMESPACE_NODE" or "ADMIN_NODE" for the
time being.
YAIM now sets both variables in this file correctly, this has always
cased a small issue of sites having to manually mount PNFS when the
"ADMIN_NODE" was not the location declared as equal to the definition
of DCACHE_ADMIN by YAIM in site-info.def.
After some delays due to side effects casued by pnfs-install.sh and
install.sh from dcache I have finally fixed my deployment testing and
am now isolating a small bug then I can declare that all confusion
should be eliminated from the next release.
The one remaining bug to establish is why dcache's install.sh is
strangely not creating the directory "/pnfs/desy.de" automatically for
clients to act as a mount point. Then once this is completed I shall
repeat the process of installing on a 4 node install until all bugs to
do with mounting PNFS and chimera have been resolved.
Then once Chimera mounting and PNFS mounting and the information system
is moderately changed I will release a new YAIM to CERN.
Once I have then tested on Solaris if no dcache release is ready I
will cut one, maybe based on my deployment changes.
Regards
Owen
